Does UnRaid OS wear out SSDs Prematurely?

[binar]

Fellow Forum Members,

Is it true the UnRaid OS and various of its docker apps eat through the Tera Byte Written (TBW) threshold of an SSD cache disk due to how it writes an excessive amount of data to the cache disk?

 

This is a concern to me because the Sandisk 4tb SSD has a 600 TBW and sells for $500. In contrast, the Samsung 860 Pro SSD costs $1000 and has a 4,800 TBW.  So the Samsung SSD has a lot longer life before it fails and is out of warranty. On the other hand, the Sandisk 4tb drive will probably burn out before its warranty runs out. 

 

Given all of that does it make more sense to just  buy a 5tb mechanical hard drive for $250 and use it as an Unraid cache disk instead of a 4tb SSD?  This way I don't have to worry about premature SSD failure which seems to be a problem due to how the UnRaid OS inflicts a lot of wear onto an SSD drive. And since the 5tb cache drive is going to work while I am asleep does it really matter it has half the data transfer speed of an SSD drive?  Or does not opting to not use an SSD drive as a cache disk defeats the entire purpose of having a cache disk in the first place?

 

I would be curious to know if anybody in the community is using a mechanical hard drive as a cache disk and if they are happy with it.  In short, I am just trying to figure out the best possible hardware to use in setting up a cache disk for my new UnRaid system.  Any opinions welcome.

 

 

 

[whipdancer]

Because I really don't fully understand the potential issues with this, I chose to make my cache a 2tb mechanical (2.5in) drive. I picked up a higher end laptop mechanical (7200rpm) drive that was well reviewed on a couple hardware sites. No issues that I've seen, but I don't actively monitor the performance of my server.

[Squid]

A small subset of users have some problems with excessive writes.  LT is aware of the issue.  AFAIK, I believe the initial thoughts on the root cause of the issue is encrypted drives (could be wrong).

 

Myself on the other hand have a 500GB 970 EVO that in roughly 6 months has accumulated ~76TB in writes.  And that is bang on where I expect it to be given my use case.  I'm not particularly worried about it. 

 

 

[binar]

Thank you Squid and Whipdancer for your posts. Drive encryption makes sense to me as the possible cause to all the excessive writing going on in some cases.  

 

Today I got on the phone with Sandisk Tech Support. They pointed me to a 4tb WD Red SA500 SSD with a higher TBW that looks promising.  The list below summarizes what I know so far:

1. The 4TB "WD Red SA500 NAS SATA SSD" has a 2500 TBW and sells for $579
2. The 4TB "Sandisk 3D NAND" has a 600 TBW and sells for $499
3. The 4TB "Samsung 860 PRO SSD" has a 4,800 TBW and sells for $1,000
4. The 4TB "Samsung 860 EVO SSD" has a 1,440 TBW and sells for $650
5. The 3.84TB "Seagate IronWolf 110 Series SSD" has a 7,000 TBW and sells for $869

If going by price and TBW alone the "Seagate IronWolf 110 Series SSD" is the best SSD you can buy to use as an UnRaid cache disk. It has a higher endurance rating than the "Samsung 860 PRO SSD" and it also costs less than the "Samsung 860 PRO SSD".  However, for some reason I don't understand it is slightly smaller than 4tb.

 

Nevertheless, spending $250 on a 5TB mechanical hard drive to use as an UnRaid cache disk still seems mighty appealing to my wallet.  For now I look forward to reading any opinions anybody is willing to share regarding this matter. 

 

 

 

[hawihoney]

Just curious: Why 4TB Cache Disk? Dockers, VMs rarely need such a huge cache. Or do you create/update 4TB array data every day?

 

As I wrote: Just curious.

 

[JonathanM]

9 hours ago, binar said:

for some reason I don't understand it is slightly smaller than 4tb.

Most likely explanation is that the actual capacity behind the wear leveling controller is the same or more than the other 4TB drives. It reserves more capacity to transparently remap bad cells and stay "defect free" for longer.

 

Many "industrial" SSD drives are odd sizes because of that.

[caplam]

For what it worth i'm concerned with the high writes. In 6 months 500TB have been written to my ssd. They should be dead (given for 300TBW).

My cache was a btrfs pool of 2 WD Blue m2 sata 500Gb. I still use one of them as xfs cache and write rate has drastically drop.

I think i'll go for single 970 pro as i may have a good deal on one 1Tb drive. But i'll wait before setting up a new btrfs pool.

To answer to the initial question i don't see an interest in a mechanical pool for cache.

The cache is interesting mainly for the speed. 

Perhaps we will have good surprises in future releases.